Citation: Cell Death and Disease (2016) 7, e2191; doi:10.1038/cddis.2016.90 OPEN & 2016 Macmillan Publishers Limited All rights reserved 2041-4889/16 www.nature.com/cddis Activated human mesenchymal stem/stromal cells suppress metastatic features of MDA-MB-231 cells by secreting IFN-β

N Yoon1,2, MS Park1,2, T Shigemoto1, G Peltier1 and RH Lee*,1

Our recent study showed that human mesenchymal stem/stromal cells (hMSCs) are activated to express (TNF)-α-related apoptosis-inducing ligand (TRAIL) by exposure to TNF-α and these activated hMSCs effectively induce apoptosis in triple-negative breast cancer MDA-MB-231 (MDA) cells in vitro and in vivo. Here, we further demonstrated that activated hMSCs not only induced apoptosis of MDA cells but also reduced metastatic features in MDA cells. These activated hMSC-exposed MDA cells showed reduced tumorigenicity and suppressed formation of lung metastasis when implanted in the mammary fat pad. Surprisingly, the activated hMSC-exposed MDA cells increased TRAIL expression, resulting in apoptosis in MDA cells. Interestingly, upregulation of TRAIL in MDA cells was mediated by -beta (IFN-β) secreted from activated hMSCs. Furthermore, IFN-β in activated hMSCs was induced by RNA and DNA released from apoptotic MDA cells in absent in melanoma 2 (AIM2) and IFN induced with helicase C domain 1 (IFIH1)-dependent manners. These observations were only seen in the TRAIL- sensitive breast cancer cell lines but not in the TRAIL-resistant breast cancer cell lines. Consistent with these results, Kaplan– Meier survival analysis also showed that lack of innate sensors detecting DNA or RNA is strongly associated with poor survival in estrogen receptor-negative breast cancer patients. In addition, cancer-associated fibroblasts (CAF) isolated from a breast cancer patient were also able to express TRAIL and IFN-β upon DNA and RNA stimulation. Therefore, our results suggest that the crosstalk between TRAIL-sensitive cancer cells and stromal cells creates a tumor-suppressive microenvironment and further provide a novel therapeutic approach to target stromal cells within cancer microenvironment for TRAIL sensitive cancer treatment. Cell Death and Disease (2016) 7, e2191; doi:10.1038/cddis.2016.90; published online 14 April 2016

Mesenchymal stem/stromal cells (MSCs) have been investi- including MDA-MB-231 (MDA) cells.23 Interestingly, TRAIL gated extensively for cancer treatment because of their expression in hMSCs is further increased by stimulation of – excellent homing ability to the tumor.1 3 However, the previous DNA and RNA released from apoptotic MDA cells and studies showed controversial results and it still remains such antitumorigenic effect of hMSCs is only shown in unclear whether MSCs promote or suppress tumor progression. TRAIL-sensitive TNBC lines.23,24 These results suggest that Many studies have shown that MSCs show pro-tumorigenic the crosstalk between hMSCs and cancer cells may differ effects by promoting proliferation of a cancer-initiating 4–7 8–10 depending on the types of cancer, and further study is population or stimulate metastasis by secreting required to examine whether the crosstalk between TRAIL- pro-tumorigenic or through crosstalk with cancer expressing activated hMSCs and TRAIL-sensitive cancer cells. Furthermore, recent studies showed that tumors recruit cells creates a tumor-suppressive environment and thereby MSCs and induce their conversion into cancer-associated – further suppresses tumor progression. fibroblasts (CAFs)11 13 that are associated with tumor – – In this study, we examined effects of activated hMSCs on progression,14 17 invasion and metastasis,16 19 therapeutic resistance15,20,21 and prognosis in breast cancer.22 metastatic features of MDA cells. Our results showed that the Our recent study demonstrated that human MSCs (hMSCs) crosstalk between TRAIL-expressing activated hMSCs and are able to express the high level of an apoptosis-inducing TRAIL-sensitive cancer cells not only induced apoptosis of factor, tumor necrosis factor (TNF)-α-related apoptosis- cancer cells but also reduced metastatic features of MDA inducing ligand (TRAIL) upon TNF-α stimulation and induce cells, which was mediated by the hMSC-derived interferon- apoptosis in triple-negative breast cancer cell (TNBC) lines beta (IFN-β).

1Texas A&M University Health Science Center, College of Medicine, Institute for Regenerative Medicine, Temple, TX, USA *Corresponding author: RH Lee, Texas A&M University Health Science Center, College of Medicine, Institute for Regenerative Medicine, 5701 Airport Road, Module C, Temple, TX 76502, USA. Tel: 254 771 6821; Fax: 254 771 6856; E-mail: [email protected] 2These authors contributed equally to this work. Abbreviations: MSCs, mesenchymal stem/stromal cells; hMSCs, human MSCs; TRAIL, tumor necrosis factor-related apoptosis-inducing ligand; rhTRAIL, recombinant human TRAIL; IFN-β, interferon-beta; rhIFN-β, recombinant human IFN-β; TNBC, triple-negative breast cancer; MDA, MDA-MB-231; ER, estrogen receptor; CAF, cancer- associated fibroblasts; act hMSCs, hMSCs activated with TNF-α; CC, coculture of cancer cells with naive hMSCs; CCT, coculture of cancer cells with hMSCs and TNF-α; poly(dA:dT), poly(deoxyadenylic-deoxythymidylic) acid; poly(I:C), polyinosinic:polycytidylic acid; PKC-α, kinase C-alpha; TLR3, Toll-like receptor 3; AIM2, absent in melanoma 2; IFIH1, interferon induced with helicase C domain 1; IRF, interferon-related factor; IFNB1, IFN-β; IS, innate sensor; CM, culture medium; sup, supernatant collected from cultured cells; FBS, fetal bovine serum; IgG, IgG control antibody Received 16.11.15; revised 24.2.16; accepted 10.3.16; Edited by M Herold IFN-β from hMSC reduces metastatic cancer features N Yoon et al 2

Results

MDA cells decrease metastatic features after coculture with TNF-α activated hMSCs. The metastatic cancer features that are characterized by high invasiveness, tumorigenicity, metastatic potential and drug resistance are closely associated with poor prognosis in several types of cancer.25 From our previous study, we demonstrated that TNF-α-activated hMSCs (act hMSCs) induce apoptosis in TRAIL-sensitive cancer cells.23 To further examine effects of act hMSCs on cancer cells, the metastatic features were analyzed in MDA cells after treatments as shown in Figure 1a. The remaining MDA cells that were isolated by magnetic-activated cell sorting (MACS) negative using CD90 (Supplementary Figure S1), a marker of hMSCs, showed the decreased expression of CD44 (Figure 1b) – a marker for cancer-initiating cells.26,27 In addition, MDA cells exhibited less migration and invasive properties after coculture with act hMSCs (Figures 1c–f). Unlike act hMSC coculture, the MDA cells from coculture with naive hMSCs (CC) or treatment of recombinant human TRAIL (rhTRAIL) increased invasion compared with the control MDA cells (Figures 1c–f). To examine tumorigenic properties of MDA cells after act hMSC coculture, the live MDA cells after act hMSC coculture were implanted into mouse mammary fat pad for the tumor burden formation. After 6 weeks of implantation, the sizes of tumor burdens formed by MDA cells from coculture were signifi- cantly smaller than naive MDA cells (Figures 1g and h). We also collected lungs of these animals to see whether any of the MDA cells metastasize from their primary tumor sites to the lung. Detection of human Alu sequences using quantita- tive PCR of genomic DNA of the lungs28 showed that only very few or no human cells were detected in the lung of animals received MDA cells isolated from coculture with act hMSCs, whereas the animals received control MDA cells showed significant numbers of metastasized cells in lungs (Figure 1i). The live MDA cells after act hMSC coculture showed decreased expression of protein kinase C-alpha (PKC-α), whereas rhTRAIL treatment increased PKC-α (Figure 1j), which is highly expressed in metastatic cancer cells.29–31 These data suggest that act hMSCs not only induce cancer cell death but also suppress metastatic features of MDA cells through coculture.

Act hMSCs induce apoptosis in rhTRAIL-resistant MDA cells. To examine the effect of act hMSCs on resistance to TRAIL-induced apoptosis in MDA cells, we treated MDA cells Figure 1 MDA cells lose their metastatic ability upon coculture with with rhTRAIL or act hMSCs as shown in Figure 2A. activated hMSCs. (a) Schematic diagram. (b) Representative images from flow – cytometry analyses detecting CD44 expression in MDA cells under different Consistent with the previous observations,32 34 rhTRAIL- conditions. Values are mean ± S.D. n = 3. (c and d) Cell migration assay; exposed MDA cells exhibited less sensitivity to the second representative images (c) and the counted number of migrated MDA cells (d)of treatment of rhTRAIL (Figure 2B(b)). We considered these migration assay. Values are mean ± S.D. n = 4; *Po0.05; **Po0.01; one-way MDA cells as rhTRAIL-resistant cells. To investigate whether ANOVA. (e and f) Invasion assay of MDA cells; representative images (e) and activated hMSCs are able to induce cell death in rhTRAIL- quantified total area at day 5 (f) of 3D invasion assay. Values are mean ± S.D. n = 6; o o resistant MDA cells, these MDA cells were cocultured with act **P 0.01; *** P 0.001; one-way ANOVA. (g and h) Images (g) and calculated volume (h) of primary tumors in the mammary fat pad of mice 6 weeks after hMSCs (Figure 2B(d)). The act hMSCs induced 470% of cell implantation of MDA cells as shown in Figure 1a. n = 5; *Po0.05; ***Po0.001; death in the rhTRAIL-resistant MDA cells (Figure 2B(d)). As a Student's t-test. (i) Metastasized MDA cells in lungs of animals from (g) using control, we treated the rhTRAIL-resistant MDA cells with quantitative PCR for human Alu sequences. n = 5; *Po0.05; Student's t-test. TNF-α, which did not induce any further apoptosis (Figure 2B (j) Western blot analysis for PKC-α expression in MDA cells isolated from different (c)). These observations suggest that the TRAIL-expressing conditions as indicated in Figure 1a

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 3

neutralizing antibody in the conditioned media, TRAIL upregulation was negated in MDA cells (Figure 3g) as well as the expression level of IRF7 (Figure 3h). Conversely, expression levels of TRAIL and IRF7 were increased in a dose-dependent manner upon recombinant human IFN-β (rhIFN-β) treatment in MDA cells (Figures 3i and j). In addition, blocking IFN-β during coculture with act hMSCs suppressed apoptosis significantly in MDA cells (Figures 3k and l). These results showed that MDA cells also expressed TRAIL during coculture with act hMSCs and the TRAIL expression in MDA cells was mediated by IFN-β in conditioned media.

DNA/RNA fragments from apoptotic cells induce secre- tion of IFN-β in act hMSCs. To examine the source of IFN-β in the conditioned media of coculture, hMSCs, and MDA cells were separated following 24-h coculture using MACS sorting. Real-time RT-PCR showed that IFN-β level was upregulated β Figure 2 Activated hMSCs induce apoptosis in rhTRAIL-resistant MDA cells. (A) in act hMSCs after coculture (Figure 4a). The IFN- was not Schematic diagram. a: MDA cells treated with rhTRAIL once and cultured for 48 h; b: detected in MDA cells in any conditions (data not shown), MDA cells treated with rhTRAIL every 24 h; c: MDA cells treated with rhTRAIL at 0 h suggesting that IFN-β is mainly secreted from act hMSCs. followed by rhTRAIL and TNF-α at 24 h; d: MDA cells treated with rhTRAIL at 0 h Type I IFNs are well-known cytokines to be produced in followed by hMSCs and TNF-α at 24 h. (B) Flow cytometry analysis for MDA cells many types of cells in response to viral infection.37 During the apoptosis after sorting from control MDA cells or coculture with activated hMSCs. innate immune response, cytosolic DNA/RNA fragments are Values are mean ± S.D. n = 3 detected as danger signals, activating several receptors and induce production of cytokines such as IFN-β.38,39 Previously, we have also shown that RNA/DNA fragments from apoptotic act hMSCs further induce apoptosis in rhTRAIL-resistant cells triggers innate sensors (ISs) such as absent in MDA cells. melanoma 2 (AIM2) and IFN induced with helicase C domain 1 (IFIH1) in hMSCs.24 Here, we also observed that IFN-β was IFN-β induces TRAIL upregulation in MDA cells during induced when hMSCs were treated with apoptotic MDA cells coculture with act hMSCs. Surprisingly, we found that MDA (Figure 4b). Messenger RNA level of IFN-β was also cells also expressed the TRAIL protein, following coculture upregulated by apoptotic MDA cell treatment (Figure 4b). with act hMSCs. Western blot analysis showed that MDA However, such upregulation was decreased upon RNase and isolated from coculture with act hMSCs also expressed a high DNase treatment (Figure 4b). Consistent with this result, IFN-β level of TRAIL protein (Figure 3a). TRAIL luciferase reporter mRNA was upregulated by co-treatment with TNF-α and either assay also confirmed the upregulation of TRAIL expression at polyinosinic:polycytidylic acid (poly(I:C)), a synthetic analog of the transcriptional level in cancer cells during coculture with double-stranded RNA, or poly(dA:dT), a repetitive synthetic act hMSCs (Figure 3b). To find out if a soluble factor induces double-stranded DNA (Figure 4c). IFN-β mRNA from hMSCs TRAIL upregulation in MDA cells, MDA cells were treated with was increased upon poly(deoxyadenylic-deoxythymidylic) the conditioned media derived from act hMSC-MDA coculture acid (poly(dA:dT)) treatment in a dose-dependent manner (CCT sup). TRAIL expression was upregulated in MDA cells (Supplementary Figure S2). To examine whether act hMSCs upon conditioned media treatment, and the upregulation was induce IFN-β secretion in an AIM2 or IFIH1-dependent negated when the conditioned media was boiled before manner, we examined expression levels of IFN-β in hMSCs treatment, suggesting soluble factors in the conditioned after AIM2 or IFIH1 siRNA transfection. Upon AIM2 or IFIH1 media may induce TRAIL upregulation (Figure 3c). blockage in hMSCs (Supplementary Figure S3), the levels of In a previous study, it was shown that TRAIL level is IFN-β were partially decreased (Figure 4d), suggesting DNA/ regulated by transcription factors related to IFNs, called IFN- RNA fragments from apoptotic cells induce secretion of IFN-β related factors (IRFs).35 Among several IRFs, Huang, et al. from hMSCs. Upregulation of AIM2, IFIH1 and IFN-β was also found that IRF1 and IRF7 upregulate TRAIL expression in observed in hMSCs upon coculture with another TRAIL- macrophages upon HIV infection. We also found that IRF7 sensitive cancer cell line, Hcc38 (Figure 4e), which is also levels in MDA cells, upon coculture with act hMSCs, were sensitive to coculture with act hMSCs.23,40 However, the upregulated markedly (Figure 3d). The upregulation of IRF7 expression levels of AIM2, IFIH1 and IFN-β in hMSCs were was also observed in the MDA cells treated with act hMSC- not upregulated following coculture with TRAIL-resistant MDA coculture conditioned media but not with boiled condi- cancer cell lines40 (Figure 4e): estrogen receptor (ER)-positive tioned media (Figure 3e). It has been shown that the breast cancer MCF-7 cells and TNBC BT20 cells.40 Further- transcription factor IRF7 is essential for the induction of IFN- more, TRAIL upregulation upon coculture with hMSCs, as α/β ,36 thus, the levels of IFN-β were assessed in observed in MDA cells, was also observed in additional conditioned media of coculture. Our result showed that IFN-β TRAIL-sensitive cancer cell line, Hcc38, but not in TRAIL- was detected only in the conditioned media from act hMSC- resistant cancer cell lines, BT20 and MCF-7 (Figure 4f). These MDA coculture (Figure 3f). When IFN-β was blocked using a results demonstrated that the IS-mediated crosstalk is

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 4

Figure 3 TRAIL is upregulated in MDA cells, following coculture with activated hMSCs. (a) Western blot analysis of TRAIL expression in MDA cells from control, TNF-α (20 ng/ml) treatment or coculture with act hMSCs (MDA+hMSCs+TNF-α; CCT). (b) Luciferase assay detecting TRAIL promoter activity in TRAIL luciferase-expressing MDA cells with following conditions; control, TNF-α (20 ng/ml) treatment, CC or CCT for 24 h. Values are mean ± S.D. (c) Quantitative RT-PCR for TRAIL from MDA cells from control, TNF-α (20 ng/ml) treatment or treated with supernatant of MDA+hMSC+TNF-α coculture (CCT sup). CCT sups were used either as it is ( − ), filtered with 0.2 μm syringe filter (Filtered) or boiled at 95 °C for 5 min (Boiled). Values are mean ± S.D. for triplicate of the assay. (d) Quantitative RT-PCRfor IRF7 from MDA cells from control, TNF-α (20 ng/ml) treatment or coculture with hMSCs (MDA+hMSCs; CC) or act hMSCs (MDA+hMSCs+TNF-α; CCT). Values are mean ± S.D. for triplicate of the assay. (e) Quantitative RT-PCRfor IRF7 from MDA cells treated as in (c). Values are mean ± S.D. for triplicate of the assay. (f) ELISA assay for IFN-β from the supernatant of MDA and hMSC coculture. Values are mean ± S.D. for octuplicate of the assay. (g and h) Quantitative RT-PCR for TRAIL (g) and IRF7 (h) from MDA cells from control or treated with supernatant of MDA+hMSC+TNF-α coculture (CCT sup) with different concentrations of IFN-β-neutralizing antibody (α-IFN-β; R&D Systems). Values are mean ± S.D. for triplicate of the assay. (i and j) Quantitative RT-PCR for TRAIL (i) and IRF7 (j) from MDA cells treated with IFN-β recombinant protein (rhIFN-β). Values are mean ± S.D. for triplicate of the assay. (k) Luciferase assay of MDA cells, following coculture with activated hMSCs and IFN-β-neutralizing antibody (α-IFN-β;20μg/ml). Values are mean ± S.D. n = 6; ****Po0.0001; Student's t-test. (l) Flow cytometry analysis of apoptosis in MDA cells, following coculture with activated hMSCs and IFN-β-neutralizing antibody (α-IFN-β;20μg/ml)

induced only between activated hMSCs and TRAIL sensitive sensor was also upregulated by rhIFN-β treatment cancer cells, suggesting that crosstalk between hMSCs and (Supplementary Figure S4B) and by treatment with the cancer cells differs depending on the types of cancer. supernatant of the coculture (Supplementary Figure S4F). AIM2 is one of the factors that are upregulated by IFN-β.41 Like AIM2, IFIH1 level was also low during coculture when We also observed that when hMSCs were treated with rhIFN-β, IFN-β-neutralizing antibody was treated (Supplementary AIM2 level was upregulated markedly (Supplementary Figure S4F). Like MDA cells, when hMSCs were treated with Figure S4A). The AIM2 level was also markedly increased in rhIFN-β, the levels of IRF7 and TRAIL were upregulated hMSCs upon treatment with the supernatant of the coculture markedly (Supplementary Figures S4C and S4D), and IFN-β with MDA cells, and such a high level of AIM2 was negated antibody negated such upregulation upon treatment with the upon IFN-β-neutralizing antibody treatment during coculture supernatant of the coculture (Supplementary Figures S4G (Supplementary Figure S4E). The level of IFIH1, the RNA and S4H). These results suggest that IS-mediated IFN-β

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 5

Figure 4 IFN-β is derived from activated hMSCs, following coculture with MDA cells. (a) Quantitative RT-PCRfor IFN-β from hMSCs after coculture with MDA (MB-231) cells. Values are mean ± S.D. for triplicate of the assay. (b) Real-time RT-PCRfor IFN-β from hMSCs treated with apoptotic MDA (MB-231) cells (Apot cells). Apoptotic MDA cells were treated with either RNase (R) or DNase (D). Values are mean ± S.D. for triplicate of the assay. (c) Quantitative RT-PCR for IFN-β from hMSCs treated with TNF-α and either poly(I:C) or poly(dA:dT) (1 μg/ml). Values are mean ± S.D. for triplicate of the assay. (d) Quantitative RT-PCR for IFN-β from hMSCs after coculture with MDA (MB-231) cells. Before coculture, hMSCs were treated with siRNA for AIM2 (siAIM2 hMSCs) or IFIH1 (siIFIH1 hMSCs). Scrambled siRNA (siSCR hMSCs) were used as a control. Values are mean ± S.D. for triplicate of the assay. (e) Quantitative RT-PCR assays for AIM2, IFIH1 and IFN-β in hMSCs, following coculture with Hcc38, MCF-7 and BT20 cells for 24 h. CC, hMSCs+cancer cells; CCT, hMSCs+cancer cells+TNF-α. Values are mean ± S.D. for triplicate of the assay. (f) Western blot assay for TRAIL expression in BT20, Hcc38 and MCF-7 cells from control (C), TNF-α (T; 20 ng/ml) treatment or coculture with act hMSCs (MDA+hMSCs+TNF-α; CCT) secretion from hMSCs creates feed-forward stimulation in β-treated MDA cells exhibited more invasive properties than both hMSCs and MDA cells, and AIM2/IFIH1 ISs and IFN-β the IgG control antibody (IgG)-treated MDA cells during are co-dependent for antitumorigenic properties of act hMSCs. coculture (Figures 5a and b). When we treated the rhIFN-β pre-treated MDA cells for 24 h as a control, the invasion was IFN-β expressed by hMSCs during coculture is one of the also reduced in a dose-dependent manner (Figure 5c). main factors that reduce metastatic features of MDA Consistent with this result, the reduced tumorigenic proper- cells. To investigate whether act hMSC-derived IFN-β is ties in MDA cells after coculture with act hMSCs were partially responsible for the reduced metastatic features of MDA cells negated by the IFN-β blockage during coculture (Figures 5d as shown in Figure 1, we applied IFN-β-neutralizing antibody and e). In addition to becoming more tumorigenic MDA cells (α-IFN-β) during coculture for 24 h and sorted MDA cells for after IFN-β blockage, metastasized MDA cell numbers in the in vitro invasion and in vivo tumorigenicity assays. The α-IFN- lungs were also increased by IFN-β blockage during

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 6

Figure 5 Activated hMSC-derived IFN-β suppress tumorigenicity and metastasis of MDA cells. (a and b) Invasion assay; representative images (a) and total area (b) of cell invasion assay of MDA cells isolated from act hMSC coculture with IgG or IFN-β-neutralizing antibody (α-IFN-β;20μg/ml). (c) Representative images of cell invasion assay of MDA cells treated with recombinant human IFN-β (rhIFN-β; 100 U/ml; R&D Systems). Values are mean ± S.D. n = 6; ***Po0.001; one-way ANOVA. (d and e) Images (d) and weight (e) of primary tumors in the mammary fat pad of mice 6 weeks after implantation of MDA cells isolated from act hMSC coculture with IgG or IFN-β-neutralizing antibody (α-IFN-β;20μg/ml). n = 6; **Po0.01; Student's t-test. (f) Metastasized MDA cells in lungs of animals from (d) using quantitative PCR detecting human Alu sequences. n = 6. (g) Western blot analysis of PKC-α phosphorylation and expression in MDA cells from act hMSC coculture with IgG or IFN-β-neutralizing antibody (α-IFN-β;20μg/ml)

coculture (Figure 5f), although there was no statistical importance of single genes or sets of genes in breast significance because of a large variation in the cell numbers. cancer.42–45 To examine prognostic significance of ISs The IFN-β-neutralizing antibody-treated MDA cells exhibited detecting DNA or RNA in breast cancer patient survival, we an increase in both phosphorylation and expression of PKC-α analyzed correlations between the expression levels of the than IgG-treated MDA cells during coculture (Figure 5g). set (AIM2, IFIH1, Toll-like receptor 3 (TLR3), IRF7, IFN- These results suggest that the crosstalk between act hMSCs β and/or TRAIL) and breast cancer patient survival rate using and MDA cells suppresses invasive and metastatic properties Kaplan–Meier survival analysis, and log- P-values were in MDA cells. calculated using the online tools.46,47 Our results showed that lack of these ISs was strongly associated with poor survival in Lack of ISs detecting DNA or RNA is strongly associated lymph node positive, ER-negative and basal-like breast with poor survival in ER-negative breast cancer patients. cancer patients (Figures 6a–c). Consistent with our in vitro Numerous studies have investigated the prognostic coculture results with ER-positive MCF-7 cells, there was no

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 7

Figure 6 Kaplan–Meier survival analysis of the ISs and related genes in breast cancer. (a-c) Kaplan–Meier survival analysis of the individual IS in all, ER-positive, ER-negative, ER-negative/lymph node positive and basal-like breast cancer patients. (d and e) Kaplan–Meier survival analysis of the gene set of ISs (AIM2, IFIH1 and TLR3), IS and IFNB1 and all (IS, IFNB1, IRF7 and TRAIL). Relapse-free survival (RFS); overall survival (OS); distant metastasis-free survival (DMFS)

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 8

Figure 7 Activated CAFs induce apoptosis in MDA cells and upregulate ISs and IFN-β upon poly(dA:dT) stimulation. (a) Flow cytometry analysis in MDA cells after coculture with activated CAFs (MDA+CAF+TNF-α). (b) Quantitative RT-PCR for AIM2, IFIH1, TLR3, TRAIL and IFN-β from CAFs treated with or without TNF-α (20 ng/ml) or poly(dA:dT) (1 μg/ml) for 24 h. Values are mean ± S.D. for triplicate of the assay. (c) Schematic diagram summarizing the observations of this study

correlation between either AIM2 or IFIH1 and ER-positive cells (Figure 7a). Furthermore, like hMSCs, CAFs were breast cancer patient survival (Supplementary Figures S4A– activated to express AIM2, IFIH1, TLR3, TRAIL and IFN-β 4C). Interestingly, the gene sets of ISs (AIM2, IFIH1 and when we activated CAFs with either poly(dA:dT) or TNF-α TLR3), ISs and IFN-β (IFNB1), and all (ISs, IFNB1, IRF7 and (Figure 7b) as observed in hMSCs. Such levels were TRAIL) showed strong correlations in ER-negative and lymph increased even more when CAFs were treated with both node positive or basal-like breast cancer patient survival poly(dA:dT) and TNF-α (Figure 4b). These results suggest (Figure 6d and e). that the signals of ISs, TRAIL and IFN-β in breast cancer sample microarrays in Figure 6 could be derived from CAFs CAFs are capable of expressing TRAIL and IFN-β. Next, and CAFs may have the ability to suppress tumor progres- we examined whether CAFs are capable of expressing sion, which could be triggered by cancer cell death. Since we TRAIL and ISs upon TNF-α stimulation and inducing IFN-β observed that CAFs can be activated to express TRAIL and expression upon the stimulation of RNA or DNA released by ISs by inflammatory stimulation such as DNA fragment (poly apoptotic cells. Our results showed that TNF-α-activated (dA:dT)) or TNF-α, cancer cell death may create the feed- CAFs (act CAF) also were able to induce cell death in MDA forward stimulation we observe here (Figure 7c).

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 9

Discussion correlate with survival and good clinical outcomes in cancer patients with hepatocellular carcinoma, with a decreased In our previous study,23,24 we showed that TRAIL-expressing risk of metastatic relapse because of the increased T-cell act hMSCs induce cell death in MDA cells, and RNA/DNA and natural killer cell infiltration.57 In addition, AIM2 and released from these apoptotic DNA cells further increase IFIH1-mediated inflammasomes may operate at the cell TRAIL expression in act hMSCs through ISs (TLR3, AIM2 autonomous level to eliminate malignant precursors through and IFIH1), resulting more cell death in MDA cells. Here, we programmed cell death or, conversely, may stimulate the showed that these RNA/DNA fragments also triggered AIM2 production of trophic factors for cancer cells and their or IFIH1-mediated IFN-β secretion in act hMSCs, which in turn stroma.58 Our analysis also showed that elevated expression induced IRF7-mediated TRAIL expression in MDA cells. levels of IFIH1 and AIM2 only correlated with good outcome in Furthermore, act hMSC-derived IFN-β also affected MDA ER-negative and basal-like breast cancers, which are mainly cells to become less aggressive and tumorigenic (Figure 7c). TRAIL-sensitive TNBC.40 Furthermore, our results showed In contrast, we showed that rhTRAIL-treated MDA cells that the hMSCs did not increase expression of ISs and IFN-β, became more invasive and resistance to TRAIL-induced following coculture with TRAIL-resistant breast cancer cell apoptosis. It has been widely observed that cancer cells that lines, MCF-7 and BT20 cells. These results suggest that survived after antitumor treatment become more aggressive, the IS-mediated feed-forward stimulation in the tumor stroma invasive, resistance to the treatment, which is a major challenge may have suppressive effects on only TRAIL-sensitive for cancer treatment. Indeed, it has been shown that TRAIL- cancer cells. sensitive cancer cells acquire TRAIL-resistant mechanisms – A recent study showed that a stroma-related gene signature upon exposure to the rhTRAIL32 34 and become an aggressive predicts clinical outcome in whole tumor samples comprising phenotype.48 Considering these problems, our finding is tumor epithelial cells and stroma.22 This study showed that the significant as it shows that activated hMSCs not only directly gene set associated with the poor outcome links to angio- induce apoptosis of MDA cells but also reduce metastatic genic, hypoxic and tumor-associated macrophage responses features in cancer cells. Furthermore, the paracrine effect of act and the gene set associated with the good outcome links hMSCs on cancer cells could lead to the new therapeutic to T-helper type 1 immune responses. However, these regimen since only 24-h coculture created the long-lasting responses of cancer microenvironment have been linked to tumor-suppressive effect on MDA cells (Figures 1 and 5). – the outcome of all types of cancers.59 62 In contrast, our Many studies have explored IFN-β as a potential cancer results show that the crosstalk signal genes between act therapy, as it reduces proliferation and inhibits tumor formation hMSCs and MDA cells accurately predicted outcome in ER- in many cancer cells.49,50 However, recombinant IFN-β protein negative and basal-like breast cancers, which are mainly treatment has been unsuccessful because of their rapid TRAIL-sensitive cancer cells. We also observed that CAFs degradation upon systemic delivery.49 To resolve such exhibited functional similarities to hMSCs, such as upregu- issue, several studies were performed to explore the lated expression of TRAIL and IFN-β upon synthetic DNA possibilities of using carriers such as hMSCs to deliver stimulation. Considering such similarities between hMSCs IFN-β by overexpressing IFN-β using viral transfection.51,52 and CAFs and strong evidences showing transition of hMSCs Here, we show, for the first time, that hMSCs have the ability to into CAFs in tumor microenvironment, we speculate that the express IFN-β naturally by crosstalk with cancer cells through initial cancer cell death induced by chemotherapy or radiation ISs, AIM2 and IFIH1, and have a tumor-suppressive role by may cause a feed-forward stimulation in the CAFs and thereby not only inducing apoptosis but also inhibiting migrating and the CAFs can suppress tumor progression or metastasis by invasive properties of cancer cells. The tumor-suppressive upregulating expression levels of TRAIL and IFN-β in TRAIL- effect of hMSC-derived IFN-β was proven by blocking IFN-β sensitive breast cancer patients. Furthermore, the crosstalk activity (Figure 5). However, the reduced tumorigenic proper- signals between cancer cells and stromal cells may help to ties in MDA cells after coculture with act hMSCs were partially predict the responses of cancer microenvironment to TRAIL- negated by the IFN-β blockage, which could be due to either sensitive cancer cells and even identify patient groups that will (i) the reduced efficacy of the neutralizing antibody over 24 h receive the greatest benefits from TRAIL or IFN-β-based or (ii) other antitumorigenic factors secreted from act hMSCs therapies. during coculture, such as DKK3 as we have shown In summary, our results suggest that the crosstalk between previously.23 Nevertheless, our result showed that IFN-β is TRAIL-sensitive cancer cells and stromal cells creates tumor- one of the major antitumorigenic factors secreted from act suppressive microenvironments and further provide a novel hMSCs by crosstalk with cancer cells. therapeutic approach to target stromal cells within cancer for Although we showed that the expression of the ISs such as TRAIL sensitive cancer treatment. AIM2 and IFIH1 was beneficial to suppress TRAIL-sensitive cancer cells, the expression of these ISs in the tumor may have a controversial effect because inflammatory responses Materials and Methods are known to increase the risk for the development and Cell preparations. hMSCs and MDA cells were prepared as previously 23 promotion of cancer.53 It has been shown that activation of described. Hcc38, MCF-7 and BT20 cells were purchased from ATCC (Manassas, VA, USA) and cultured as suggested by the manufacturer. Cells were passaged TLR3 in melanoma cancer cells increases their tumorigenicity 54 1 : 3 when 70% confluent. CAFs were purchased from Asterand Bioscience (Detroit, and migration capacity. In contrast, expression of functional MI, USA), and cultured using α-MEM containing 16% fetal bovine serum (FBS; lot- TLR3 in cancer cells reduces tumorigenicity and exhibits selected for rapid growth of hMSCs; Atlanta Biologicals, Inc., Norcross, GA, USA), 55,56 apoptosis. Also, intra-tumoral expression of TLR3 100 units/ml penicillin, 100 μg/ml streptomycin and 2 mM L-glutamine (Life

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 10

Technologies, Carlsbad, CA, USA). Cells were passaged 1 : 3 when 70% confluent. Cell death assay. hMSCs and MDA cells from 24-h coculture experiments with Passages 2 to 3 of hMSCs and passages 4 to 5 of CAF were used for all or without TNF-α (20 ng/ml) were labeled with CD90 and incubated at room experiments. temperature for 20 min followed by 300 ng/ml Annexin V (Annexin V-FITC Apoptosis Detection Kit; Sigma-Aldrich, St. Louis, MO, USA) and 4 μg/ml 7-aminoactinomycin Magnetic-activated cell sorting. hMSCs and cancer cells (MDA or MCF-7 D (Sigma-Aldrich) to detect apoptotic cells. The labeled cells were analyzed by flow cells) were cocultured as previously described.23 Briefly, hMSCs (1 × 105 cells) and cytometry for cell death assay. cancer cells (1 × 105 cells) were plated in a six-well plate with or without TNF-α (20 ng/ml). After 24 h of coculture, the cells were labeled with CD90-PE (clone RNA extraction from cultured cells and real-time RT-PCR Thy1/310; Beckman Coulter, Brea, CA, USA) for 30 min at 4 °C and anti-PE analysis. hMSCs, MDA, Hcc38, MCF-7 and BT20 cells were separated by MicroBeads (Miltenyl Biotec, Bergisch Gladbach, Germany) for 30 min at 4 °C and either MACS LS or LD columns after coculture. RNA was extracted using RNeasy separated using LS (positive separation for hMSCs) or LD (negative separations for Mini Kit (QIAGEN). Real-time RT-PCR analyses were performed as previously cancer cells) columns on QuadroMACS separator (Miltenyl Biotec) according to the described.23 List of primers and probes is in the Supplementary Information. manufacturer’s protocol. The purity of sorted cells was checked using flow cytometry (Cytomics FC500; Beckman Coulter). Transfections with siRNA. hMSCs (5 × 104 cells per well in six-well plate) Flow cytometry analysis. hMSCs and MDA cells from coculture were were transfected by incubating 4 h with 20 nM siRNA for AIM2, IFIH1 or scrambled labeled with anti-CD90-PE (clone Thy1/310; Beckman Coulter) and CD44-PE/Cy5 siRNA A (Santa Cruz Biotechnology, Dallas, TX, USA) using Lipofectamine (clone G44-26; BD Biosciences) for 45 min on ice, washed with PBS by RNAiMax reagent (Life Technologies). Following transfection, the cells were centrifugation and analyzed using flow cytometry. recovered with 16% FBS containing α-MEM for 2–4 h and then cocultured with MDA cells. Duplicate wells were also prepared to check the transfection efficiency Cell migration assay. In vitro cell migration was examined using Transwell using real-time RT-PCR assay. culture inserts (BD Biosciences, Franklin Lakes, NJ, USA) with 8 μm pore filter 5 inserts for 24-well plates. Briefly, 5 × 10 cells suspended in serum-free DMEM were Western blot analysis. Western blot analyses were performed after sorting added to the inserts, the upper chamber, which was placed in a 24-well culture MDA cells or hMSCs from coculture, as described above. Procedures for western plate. FBS was added to the lower chamber at the final concentration of 2% as a blot analyses were described previously.23 Antibodies used in this study are listed in chemoattractant. After 6 h, the upper insert was removed, washed and the non- Supplementary Information. filtered cells were gently removed with a cotton swab. Filtered cells located on the lower side of the chamber were stained with crystal violet, photographed and counted using Image J ver.1.48 (http://imagej.nih.gov/ij/). Luciferase assay. MDA cells that were stably transfected with pGL4 reporting vector (Promega, Madison, WI, USA) containing TRAIL promoter (see Cell invasion assay. In vitro cell invasion assay was performed using Cultrex Supplementary Information for plasmid construction and stable transfection) were 96 Well 3D Spheroid BME Cell Invasion (Trevigen, Inc., Gaithersburg, MD, USA) cocultured with hMSCs (10 000 cells per well in 96-well plate) with or without according to the manufacturer’s protocol. Briefly, 3 000 cells resuspended in TNF-α. After 24 h, luciferase activity was detected to read TRAIL expression levels spheroid formation ECM were dispensed in a well of the 3D Culture Qualified 96 using Bright-Glo Luciferase Assay System (Promega) according to the Well Spheroid Formation Plate. The plate was then spun down and incubated at manufacturer's protocol. In parallel, these MDA cells were cocultured with hMSCs 37 °C for spheroid formation. After 24 h, invasion matrix was added and incubated in a six-well plate for apoptosis assay. Luciferase activity levels were normalized to at 37 °C. At chosen time points, images of invading spheroids were taken, and the the percentage of live MDA cells. total area of invading spheroid was calculated with Image J.

In vivo tumorigenicity and metastasis assay. Either control MDA cells Conflict of Interest (1 × 104 or 1 × 105 in 100 μl of HBSS) or MDA cells that were isolated from The authors declare no conflict of interest. coculture with activated hMSCs and IgG or IFN-β-neutralizing antibody were inoculated into mammary fat pad of 6-week-old female NOD/SCID mice. After the 1. Kanehira M, Xin H, Hoshino K, Maemondo M, Mizuguchi H, Hayakawa T et al. Targeted single injection of MDA cells, mice were palpated for tumor growth weekly after delivery of NK4 to multiple lung tumors by bone marrow-derived mesenchymal stem cells. tumor implantation. Once palpable, tumors were measured in two dimensions Cancer Gene Ther 2007; 14: 894–903. (length and width) using a digital caliper to calculate the volume. After 4–6 weeks 2. Kucerova L, Altanerova V, Matuskova M, Tyciakova S, Altaner C. Adipose tissue-derived from cell inoculation, all mice were killed, and tumors in the fat pad were collected, human mesenchymal stem cells mediated prodrug cancer gene therapy. Cancer Res 2007; – and tumor volumes were measured. To examine metastasis, lungs were collected 67: 6304 6313. 3. Studeny M, Marini FC, Champlin RE, Zompetta C, Fidler IJ, Andreeff M. Bone marrow- for genomic DNA isolation to detect hAlu sequences using real-time PCR.23 derived mesenchymal stem cells as vehicles for interferon-beta delivery into tumors. Cancer Res 2002; 62: 3603–3608. Animals. Six-week-old female NOD/SCID mice (NOD.CB17-Prkdcscid/J) from 4. Hsu HS, Lin JH, Hsu TW, Su K, Wang CW, Yang KY et al. Mesenchymal stem cells enhance the Jackson Laboratory (Bar Harbor, ME, USA) were used under a protocol lung cancer initiation through activation of IL-6/JAK2/STAT3 pathway. Lung Cancer 2012; 75: approved by the Institutional Animal Care and Use Committee of Texas A&M Health 167–177. Science Center College of Medicine. 5. Harper ME, Antoniou A, Villalobos-Menuey E, Russo A, Trauger R, Vendemelio M et al. Characterization of a novel metabolic strategy used by drug-resistant tumor cells. FASEB J – Genomic DNA extraction/real-time PCR assays for human Alu 2002; 16: 1550 1557. 6. Liu S, Ginestier C, Ou SJ, Clouthier SG, Patel SH, Monville F et al. Breast cancer stem cells sequences. Genomic DNA from lung samples was extracted, and real-time 28 are regulated by mesenchymal stem cells through networks. Cancer Res 2011; 71: PCR was performed to detect human Alu signals, as previously described. 614–624. 7. Tsai KS, Yang SH, Lei YP, Tsai CC, Chen HW, Hsu CY et al. Mesenchymal stem cells Preparation of apoptotic MDA cells. MDA cells were plated in serum-free promote formation of colorectal tumors in mice. Gastroenterology 2011; 141: 1046–1056. α-MEM with 100 ng/ml rhTRAIL (R&D Systems, Minneapolis, MN, USA). After 24 h 8. Swamydas M, Ricci K, Rego SL, Dreau D. Mesenchymal stem cell-derived CCL-9 and at 37 °C, floating cells were collected and washed by centrifugation at 500 x g for CCL-5 promote mammary tumor cell invasion and the activation of matrix metalloprotei- 5 min. The pellet was resuspended in 2% culture medium (2% CM; alpha-MEM nases. Cell Adh Migr 2013; 7: 315–324. 9. Mi Z, Bhattacharya SD, Kim VM, Guo H, Talbot LJ, Kuo PC. promotes CCL5- containing 2% FBS, 100 units/ml penicillin, 100 μg/ml streptomycin and 2 mM L- mesenchymal stromal cell-mediated breast cancer metastasis. Carcinogenesis 2011; 32: glutamine) and plated on hMSC containing wells. For RNase and DNase treatment, 477–487. apoptotic MDA cells were washed by centrifugation, resuspended in 0.2 ml PBS 10. Karnoub AE, Dash AB, Vo AP, Sullivan A, Brooks MW, Bell GW et al. Mesenchymal stem μ containing either 20 g of RNase (QIAGEN, Valencia, CA, USA) or 30 units of cells within tumour stroma promote breast cancer metastasis. Nature 2007; 449:557–563. DNase (QIAGEN), and incubated for 2 h at 37 °C. The cells were washed by 11. Jung Y, Kim JK, Shiozawa Y, Wang J, Mishra A, Joseph J et al. Recruitment of mesenchymal centrifugation and resuspended in 2% CM before adding to hMSC containing wells. stem cells into prostate tumours promotes metastasis. Nat Commun 2013; 4: 1795.

Cell Death and Disease IFN-β from hMSC reduces metastatic cancer features N Yoon et al 11

12. Mishra PJ, Mishra PJ, Humeniuk R, Medina DJ, Alexe G, Mesirov JP et al. Carcinoma- 39. Takaoka A, Taniguchi T, Cytosolic DNA. recognition for triggering innate immune responses. associated fibroblast-like differentiation of human mesenchymal stem cells. Cancer Res Adv Drug Deliv Rev 2008; 60: 847–857. 2008; 68: 4331–4339. 40. Rahman M, Davis SR, Pumphrey JG, Bao J, Nau MM, Meltzer PS et al. TRAIL induces 13. Spaeth EL, Dembinski JL, Sasser AK, Watson K, Klopp A, Hall B et al. Mesenchymal stem apoptosis in triple-negative breast cancer cells with a mesenchymal phenotype. Breast cell transition to tumor-associated fibroblasts contributes to fibrovascular network expansion Cancer Res Treat 2009; 113: 217–230. and tumor progression. PLoS One 2009; 4: e4992. 41. Fang R, Hara H, Sakai S, Hernandez-Cuellar E, Mitsuyama M, Kawamura I et al. Type I 14. Olumi AF, Grossfeld GD, Hayward SW, Carroll PR, Tlsty TD, Cunha GR. Carcinoma- interferon signaling regulates activation of the absent in melanoma 2 inflammasome during associated fibroblasts direct tumor progression of initiated human prostatic epithelium. Streptococcus pneumoniae infection. Infect Immun 2014; 82: 2310–2317. Cancer Res 1999; 59: 5002–5011. 42. van 't Veer LJ, Dai H, van de Vijver MJ, He YD, Hart AA, Mao M et al. Gene expression 15. Kharaishvili G, Simkova D, Bouchalova K, Gachechiladze M, Narsia N, Bouchal J. The role profiling predicts clinical outcome of breast cancer. Nature 2002; 415: 530–536. of cancer-associated fibroblasts, solid stress and other microenvironmental factors in tumor 43. van de Vijver MJ, He YD, van't Veer LJ, Dai H, Hart AA, Voskuil DW et al. A gene-expression progression and therapy resistance. Cancer Cell Int 2014; 14:41–48. signature as a predictor of survival in breast cancer. N Engl J Med 2002; 347: 1999–2009. 16. Mao Y, Keller ET, Garfield DH, Shen K, Wang J. Stromal cells in tumor microenvironment 44. Wang Y, Klijn JG, Zhang Y, Sieuwerts AM, Look MP, Yang F et al. Gene-expression profiles and breast cancer. Cancer Metastasis Rev 2013; 32: 303–315. to predict distant metastasis of lymph-node-negative primary breast cancer. Lancet 2005; 17. Elenbaas B, Weinberg RA. Heterotypic signaling between epithelial tumor cells and 365: 671–679. fibroblasts in carcinoma formation. Exp Cell Res 2001; 264: 169–184. 45. Teschendorff AE, Caldas C. A robust classifier of high predictive value to identify good 18. Karagiannis GS, Poutahidis T, Erdman SE, Kirsch R, Riddell RH, Diamandis EP. Cancer- prognosis patients in ER-negative breast cancer. Breast Cancer Res 2008; 10: R73. associated fibroblasts drive the progression of metastasis through both paracrine and 46. Ringner M, Fredlund E, Hakkinen J, Borg A, Staaf J. GOBO: gene expression-based mechanical pressure on cancer tissue. Mol Cancer Res 2012; 10: 1403–1418. outcome for breast cancer online. PLoS One 2011; 6: e17911. 19. Tommelein J, Verset L, Boterberg T, Demetter P, Bracke M, De Wever O. Cancer-associated 47. Gyorffy B, Lanczky A, Eklund AC, Denkert C, Budczies J, Li Q et al. An online fibroblasts connect metastasis-promoting communication in colorectal cancer. Front Oncol survival analysis tool to rapidly assess the effect of 22,277 genes on breast cancer 2015; 5: 63. prognosis using microarray data of 1,809 patients. Breast Cancer Res Treat 2010; 123: 20. Kinugasa Y, Matsui T, Takakura N. CD44 expressed on cancer-associated fibroblasts is a 725–731. functional molecule supporting the stemness and drug resistance of malignant cancer cells 48. Jeon YJ, Middleton J, Kim T, Lagana A, Piovan C, Secchiero P et al. A set of NF-kappaB- in the tumor microenvironment. Stem Cells 2014; 32: 145–156. regulated microRNAs induces acquired TRAIL resistance in lung cancer. Proc Natl Acad Sci 21. Micke P, Ostman A. Exploring the tumour environment: cancer-associated fibroblasts as USA 2015; 112: E3355–E3364. targets in cancer therapy. Expert Opin Ther Targets 2005; 9: 1217–1233. 49. Qin XQ, Tao N, Dergay A, Moy P, Fawell S, Davis A et al. Interferon-beta gene therapy 22. Finak G, Bertos N, Pepin F, Sadekova S, Souleimanova M, Zhao H et al. Stromal gene inhibits tumor formation and causes regression of established tumors in immune- expression predicts clinical outcome in breast cancer. Nat Med 2008; 14: 518–527. deficient mice. Proc Natl Acad Sci USA 1998; 95: 14411–14416. 23. Lee RH, Yoon N, Reneau JC, Prockop DJ. Preactivation of human MSCs with TNF-alpha 50. Johns TG, Mackay IR, Callister KA, Hertzog PJ, Devenish RJ, Linnane AW. Antiproliferative enhances tumor-suppressive activity. Cell Stem Cell 2012; 11: 825–835. potencies of on melanoma cell lines and xenografts: higher efficacy of 24. Yoon N, Park MS, Peltier GC, Lee RH. Pre-activated human mesenchymal stromal cells in interferon beta. J Natl Cancer Inst 1992; 84: 1185–1190. combination with doxorubicin synergistically enhance tumor-suppressive activity in mice. 51. Matsuzuka T, Rachakatla RS, Doi C, Maurya DK, Ohta N, Kawabata A et al. Human Cytotherapy 2015; 17: 1332–1341. umbilical cord matrix-derived stem cells expressing interferon-beta gene significantly 25. Fidler IJ. Biology of cancer metastasis. In: Abeloff MD, Armitage JO, Niederhuber JE, Kastan attenuate bronchioloalveolar carcinoma xenografts in SCID mice. Lung Cancer 2010; 70: MB, McKenna WG (eds). Clinical Oncology, 3rd edn. Elsevier Science: Philadelphia, PA, 28–36. USA, 2004, pp 59–79. 52. Wang GX, Zhan YA, Hu HL, Wang Y, Fu B. Mesenchymal stem cells modified to express 26. Al-Hajj M, Wicha MS, Benito-Hernandez A, Morrison SJ, Clarke MF. Prospective identification interferon-beta inhibit the growth of prostate cancer in a mouse model. J Int Med Res 2012; of tumorigenic breast cancer cells. Proc Natl Acad Sci USA 2003; 100:3983–3988. 40: 317–327. 27. Reim F, Dombrowski Y, Ritter C, Buttmann M, Hausler S, Ossadnik M et al. Immunoselection 53. Mantovani A. Cancer: inflammation by remote control. Nature 2005; 435: 752–753. of breast and ovarian cancer cells with trastuzumab and natural killer cells: selective escape 54. Goto Y, Arigami T, Kitago M, Nguyen SL, Narita N, Ferrone S et al. Activation of Toll-like of CD44high/CD24low/HER2low breast cancer stem cells. Cancer Res 2009; 69: receptors 2, 3, and 4 on human melanoma cells induces inflammatory factors. Mol Cancer 8058–8066. Ther 2008; 7: 3642–3653. 28. Lee RH, Pulin AA, Seo MJ, Kota DJ, Ylostalo J, Larson BL et al. Intravenous hMSCs improve 55. Salaun B, Lebecque S, Matikainen S, Rimoldi D, Romero P. Toll-like receptor 3 myocardial infarction in mice because cells embolized in lung are activated to secrete the expressed by melanoma cells as a target for therapy? Clin Cancer Res 2007; 13(15 Pt 1): anti-inflammatory protein TSG-6. Cell Stem Cell 2009; 5:54–63. 4565–4574. 29. Kim J, Thorne SH, Sun L, Huang B, Mochly-Rosen D. Sustained inhibition of PKCalpha 56. Salaun B, Coste I, Rissoan MC, Lebecque SJ, Renno T. TLR3 can directly trigger apoptosis reduces intravasation and lung seeding during mammary tumor metastasis in an in vivo in human cancer cells. J Immunol 2006; 176: 4894–4901. mouse model. Oncogene 2011; 30: 323–333. 57. Chew V, Abastado JP. Immunomodulation of the tumor microenvironment by Toll-like 30. Byers HR, Boissel SJ, Tu C, Park HY. RNAi-mediated knockdown of protein kinase C-alpha receptor-3 (TLR3) ligands. Oncoimmunology 2013; 2: e23493. inhibits cell migration in MM-RU human metastatic melanoma cell line. Melanoma Res 2010; 58. Zitvogel L, Kepp O, Galluzzi L, Kroemer G. Inflammasomes in carcinogenesis and anticancer 20: 171–178. immune responses. Nat Immunol 2012; 13: 343–351. 31. La Porta CA, Di Dio A, Comolli R. Inhibition of PKCalpha decreases the gelatinase activity 59. Ellis LM, Hicklin DJ. VEGF-targeted therapy: mechanisms of anti-tumour activity. Nat Rev and the angiogenic and metastatic ability of the highly metastatic B16 murine Cancer 2008; 8: 579–591. melanoma cells. Angiogenesis 1999; 3: 241–247. 60. Wilson WR, Hay MP. Targeting hypoxia in cancer therapy. Nat Rev Cancer 2011; 11: 32. Yoshida T, Zhang Y, Rivera Rosado LA, Zhang B. Repeated treatment with subtoxic doses of 393–410. TRAIL induces resistance to apoptosis through its death receptors in MDA-MB-231 breast 61. De Palma M, Lewis CE. Macrophage regulation of tumor responses to anticancer therapies. cancer cells. Mol Cancer Res 2009; 7: 1835–1844. Cancer Cell 2013; 23: 277–286. 33. Spencer SL, Gaudet S, Albeck JG, Burke JM, Sorger PK. Non-genetic origins of cell-to-cell 62. Disis ML. Immune regulation of cancer. J Clin Oncol 2010; 28:4531–4538. variability in TRAIL-induced apoptosis. Nature 2009; 459: 428–432. 34. Lane D, Cote M, Grondin R, Couture MC, Piche A. Acquired resistance to TRAIL-induced apoptosis in human ovarian cancer cells is conferred by increased turnover of mature Cell Death and Disease is an open-access journal caspase-3. Mol Cancer Ther 2006; 5: 509–521. published by Nature Publishing Group. This work is 35. Huang Y, Walstrom A, Zhang L, Zhao Y, Cui M, Ye L et al. Type I interferons and interferon licensed under a Creative Commons Attribution 4.0 International regulatory factors regulate TNF-related apoptosis-inducing ligand (TRAIL) in HIV-1-infected macrophages. PLoS One 2009; 4: e5397. License. The images or other third party material in this article are 36. Honda K, Yanai H, Negishi H, Asagiri M, Sato M, Mizutani T et al. IRF-7 is the master included in the article’s Creative Commons license, unless indicated regulator of type-I interferon-dependent immune responses. Nature 2005; 434: 772–777. otherwise in the credit line; if the material is not included under the 37. Perry AK, Chen G, Zheng D, Tang H, Cheng G. The host type I interferon response to viral Creative Commons license, users will need to obtain permission from and bacterial infections. Cell Res 2005; 15: 407–422. 38. Unterholzner L, Keating SE, Baran M, Horan KA, Jensen SB, Sharma S et al. IFI16 is an the license holder to reproduce the material. To view a copy of this innate immune sensor for intracellular DNA. Nat Immunol 2010; 11: 997–1004. license, visit http://creativecommons.org/licenses/by/4.0/

Supplementary Information accompanies this paper on Cell Death and Disease website (http://www.nature.com/cddis)

Cell Death and Disease